{"id":"https://openalex.org/W7125492685","doi":"https://doi.org/10.1162/tacl.a.599","title":"On the Limitations of Language-targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning","display_name":"On the Limitations of Language-targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning","publication_year":2026,"publication_date":"2026-01-16","ids":{"openalex":"https://openalex.org/W7125492685","doi":"https://doi.org/10.1162/tacl.a.599"},"language":"en","primary_location":{"id":"doi:10.1162/tacl.a.599","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.599","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/TACL.a.599/2577928/tacl.a.599.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/TACL.a.599/2577928/tacl.a.599.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5123680050","display_name":"Simon Kurz","orcid":null},"institutions":[{"id":"https://openalex.org/I200332995","display_name":"TU Dortmund University","ror":"https://ror.org/01k97gp34","country_code":"DE","type":"education","lineage":["https://openalex.org/I200332995"]},{"id":"https://openalex.org/I4387154616","display_name":"Lamarr Institute for Machine Learning and Artificial Intelligence","ror":"https://ror.org/04s11ea33","country_code":null,"type":"facility","lineage":["https://openalex.org/I135140700","https://openalex.org/I200332995","https://openalex.org/I4210098062","https://openalex.org/I4210144576","https://openalex.org/I4387154616","https://openalex.org/I4923324","https://openalex.org/I4923324"]}],"countries":["DE"],"is_corresponding":true,"raw_author_name":"Simon Kurz","raw_affiliation_strings":["Department of Computer Science, TU Dortmund University, Germany","Lamarr Institute for Machine Learning and Artificial Intelligence, Germany"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, TU Dortmund University, Germany","institution_ids":["https://openalex.org/I200332995"]},{"raw_affiliation_string":"Lamarr Institute for Machine Learning and Artificial Intelligence, Germany","institution_ids":["https://openalex.org/I4387154616"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5107997940","display_name":"Jian-Jia Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I200332995","display_name":"TU Dortmund University","ror":"https://ror.org/01k97gp34","country_code":"DE","type":"education","lineage":["https://openalex.org/I200332995"]},{"id":"https://openalex.org/I4387154616","display_name":"Lamarr Institute for Machine Learning and Artificial Intelligence","ror":"https://ror.org/04s11ea33","country_code":null,"type":"facility","lineage":["https://openalex.org/I135140700","https://openalex.org/I200332995","https://openalex.org/I4210098062","https://openalex.org/I4210144576","https://openalex.org/I4387154616","https://openalex.org/I4923324","https://openalex.org/I4923324"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Jian-Jia Chen","raw_affiliation_strings":["Department of Computer Science, TU Dortmund University, Germany","Lamarr Institute for Machine Learning and Artificial Intelligence, Germany"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science, TU Dortmund University, Germany","institution_ids":["https://openalex.org/I200332995"]},{"raw_affiliation_string":"Lamarr Institute for Machine Learning and Artificial Intelligence, Germany","institution_ids":["https://openalex.org/I4387154616"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120003583","display_name":"Lucie Flek","orcid":null},"institutions":[{"id":"https://openalex.org/I135140700","display_name":"University of Bonn","ror":"https://ror.org/041nas322","country_code":"DE","type":"education","lineage":["https://openalex.org/I135140700"]},{"id":"https://openalex.org/I4210164149","display_name":"Bonn Aachen International Center for Information Technology","ror":"https://ror.org/054zhq066","country_code":"DE","type":"facility","lineage":["https://openalex.org/I135140700","https://openalex.org/I155417937","https://openalex.org/I4210164149","https://openalex.org/I4923324"]},{"id":"https://openalex.org/I4387154616","display_name":"Lamarr Institute for Machine Learning and Artificial Intelligence","ror":"https://ror.org/04s11ea33","country_code":null,"type":"facility","lineage":["https://openalex.org/I135140700","https://openalex.org/I200332995","https://openalex.org/I4210098062","https://openalex.org/I4210144576","https://openalex.org/I4387154616","https://openalex.org/I4923324","https://openalex.org/I4923324"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Lucie Flek","raw_affiliation_strings":["Bonn-Aachen International Center for Information Technology, University of Bonn, Germany","Lamarr Institute for Machine Learning and Artificial Intelligence, Germany"],"affiliations":[{"raw_affiliation_string":"Bonn-Aachen International Center for Information Technology, University of Bonn, Germany","institution_ids":["https://openalex.org/I4210164149","https://openalex.org/I135140700"]},{"raw_affiliation_string":"Lamarr Institute for Machine Learning and Artificial Intelligence, Germany","institution_ids":["https://openalex.org/I4387154616"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5123637192","display_name":"Zhixue Zhao","orcid":null},"institutions":[{"id":"https://openalex.org/I91136226","display_name":"University of Sheffield","ror":"https://ror.org/05krs5044","country_code":"GB","type":"education","lineage":["https://openalex.org/I91136226"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Zhixue Zhao","raw_affiliation_strings":["Computer Science School, University of Sheffield, United Kingdom. zhixue.zhao@sheffield.ac.uk"],"affiliations":[{"raw_affiliation_string":"Computer Science School, University of Sheffield, United Kingdom. zhixue.zhao@sheffield.ac.uk","institution_ids":["https://openalex.org/I91136226"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5123680050"],"corresponding_institution_ids":["https://openalex.org/I200332995","https://openalex.org/I4387154616"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.16894172,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"14","issue":null,"first_page":"167","last_page":"192"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.23119999468326569,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.23119999468326569,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.15479999780654907,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.12250000238418579,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.9470999836921692},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.8413000106811523},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.675000011920929},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.5723000168800354},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5559999942779541},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.4948999881744385}],"concepts":[{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.9470999836921692},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8618999719619751},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.8413000106811523},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.675000011920929},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.6305000185966492},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6026999950408936},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.5723000168800354},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5559999942779541},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.4948999881744385},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4643999934196472},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.38960000872612},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.3785000145435333},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.35499998927116394},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2874000072479248},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.27970001101493835}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1162/tacl.a.599","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.599","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/TACL.a.599/2577928/tacl.a.599.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},{"id":"pmh:oai:eprints.whiterose.ac.uk:239669","is_oa":false,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4306400854","display_name":"White Rose Research Online (University of Leeds, The University of Sheffield, University of York)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I2800616092","host_organization_name":"White Rose University Consortium","host_organization_lineage":["https://openalex.org/I2800616092"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":null,"raw_type":"PeerReviewed"}],"best_oa_location":{"id":"doi:10.1162/tacl.a.599","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.599","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/TACL.a.599/2577928/tacl.a.599.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.8354071974754333,"display_name":"Quality Education"}],"awards":[{"id":"https://openalex.org/G6955755495","display_name":null,"funder_award_id":"Germany","funder_id":"https://openalex.org/F4320321114","funder_display_name":"Bundesministerium f\u00fcr Bildung und Forschung"}],"funders":[{"id":"https://openalex.org/F4320320342","display_name":"University of Sheffield","ror":"https://ror.org/05krs5044"},{"id":"https://openalex.org/F4320321114","display_name":"Bundesministerium f\u00fcr Bildung und Forschung","ror":"https://ror.org/04pz7b180"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7125492685.pdf","grobid_xml":"https://content.openalex.org/works/W7125492685.grobid-xml"},"referenced_works_count":30,"referenced_works":["https://openalex.org/W2912924812","https://openalex.org/W2946609015","https://openalex.org/W2979826702","https://openalex.org/W3045958725","https://openalex.org/W3106889297","https://openalex.org/W3137147200","https://openalex.org/W3177450194","https://openalex.org/W4380346028","https://openalex.org/W4385572966","https://openalex.org/W4385574361","https://openalex.org/W4389520538","https://openalex.org/W4389524039","https://openalex.org/W4399051247","https://openalex.org/W4401042750","https://openalex.org/W4402640274","https://openalex.org/W4402667037","https://openalex.org/W4402670014","https://openalex.org/W4402671717","https://openalex.org/W4402671893","https://openalex.org/W4404781711","https://openalex.org/W4404782158","https://openalex.org/W4404782843","https://openalex.org/W4404782943","https://openalex.org/W4404918643","https://openalex.org/W4409157490","https://openalex.org/W4412886720","https://openalex.org/W4412888554","https://openalex.org/W4412888769","https://openalex.org/W4416035478","https://openalex.org/W6893007900"],"related_works":[],"abstract_inverted_index":{"Abstract":[0],"Recent":[1],"advances":[2],"in":[3,15,46],"large":[4],"language":[5,67,119],"model":[6],"(LLM)":[7],"pruning":[8,65,84,94,102,150],"have":[9],"shown":[10],"state-of-the-art":[11],"(SotA)":[12],"compression":[13],"results":[14,111],"post-training":[16],"and":[17,42,59,92,104,123,178],"retraining-free":[18],"settings":[19],"while":[20,114],"maintaining":[21],"high":[22,125],"predictive":[23],"performance.":[24,135],"However,":[25],"previous":[26],"research":[27],"mainly":[28],"considered":[29],"calibrating":[30],"based":[31],"on":[32,116],"English":[33],"text,":[34],"despite":[35],"the":[36,57,74,99,117,166],"multilingual":[37,66,85],"nature":[38],"of":[39,56,138,148,168],"modern":[40],"LLMs":[41],"their":[43],"frequent":[44],"use":[45],"non-English":[47],"languages.":[48],"This":[49],"analysis":[50,137],"paper":[51],"conducts":[52],"an":[53],"in-depth":[54],"investigation":[55],"performance":[58],"internal":[60,139],"representation":[61],"changes":[62],"associated":[63],"with":[64],"models":[68,86],"for":[69,83,175],"monolingual":[70],"applications.":[71],"We":[72,96],"present":[73],"first":[75],"comprehensive":[76],"empirical":[77],"study,":[78],"comparing":[79],"different":[80,143],"calibration":[81,115],"languages":[82],"across":[87],"diverse":[88],"languages,":[89],"tasks,":[90],"models,":[91],"SotA":[93],"techniques.":[95],"further":[97],"analyze":[98],"latent":[100],"subspaces,":[101],"masks,":[103],"individual":[105],"neurons":[106],"within":[107],"pruned":[108],"models.":[109],"Our":[110],"reveal":[112],"that":[113,172],"target":[118],"effectively":[120,154],"retains":[121],"perplexity":[122],"yields":[124],"signal-to-noise":[126],"ratios,":[127],"it":[128],"does":[129],"not":[130],"consistently":[131],"improve":[132],"downstream":[133],"task":[134],"Further":[136],"representations":[140],"at":[141],"three":[142],"levels":[144],"highlights":[145],"broader":[146],"limitations":[147],"current":[149],"approaches:":[151],"While":[152],"they":[153],"preserve":[155],"dominant":[156],"information":[157],"like":[158],"language-specific":[159],"features,":[160],"this":[161],"is":[162],"insufficient":[163],"to":[164],"counteract":[165],"loss":[167],"nuanced,":[169],"language-agnostic":[170],"features":[171],"are":[173],"crucial":[174],"knowledge":[176],"retention":[177],"reasoning.":[179]},"counts_by_year":[],"updated_date":"2026-04-10T15:06:20.359241","created_date":"2026-01-24T00:00:00"}